**********************************************************************************************
* Reproduces Table A3
*
* Ajzenman, Elacqua, Jaimovich and Pérez-Núñez (2024)."Humans versus Chatbots: Scaling-up behavioral interventions to reduce teacher shortages"  	  		
*
**********************************************************************************************


 *The initial bot was activated on 11/30/2021, identified as 22614. Exclude individuals who received human interventions prior to this date.
 
preserve
drop if fecha_date<22614 & grupo_gen==1


///// Education Outcomes

*Preferences and Enrollment in Educational Programs: Human and Bot Effects. Only students who were exposed to the treatment during weeks when there was overlap in the timing of both human and bot interventions.

reg education_rankfirst i.grupo_gen education_preference sexo i.dependencia_colegio top_30,r
estadd ysumm 
outreg2 using "${output}/TableA2.tex", replace sdec(4) tex  keep(i.grupo_gen)  addstat(Mean of dependent variable, e(ymean)) label nocons

foreach var in education_proportion education_listed d_enroll_education{
	
reg `var' i.grupo_gen education_preference sexo i.dependencia_colegio top_30,r
estadd ysumm 
outreg2 using "${output}/TableA2.tex", append sdec(4) tex  keep(i.grupo_gen)  addstat(Mean of dependent variable, e(ymean)) label nocons
eststo clear
}

*Human vs Bot effect differences p-value
foreach var in education_rankfirst education_proportion education_listed d_enroll_education{
	
reg `var' i.grupo_gen education_preference sexo i.dependencia_colegio top_30,r
test 1.grupo_gen=2.grupo_gen	
}
restore 

*Note: Calculations of the significant difference in the bot vs human effect were manually added to the table according to the results produced by the commands run earlier in this code.